sixhobbits

I've been running k3 on a single bigger hetzner dedicated machine to power https://demo.opensuite.online, a work-in-progress open source replacement for Google Workspace, a distribution of what the Dutch government is trying to switch to but they seemingly don't care about UX at all so I wanted to try build on their foundations (which in turn are basically just nextcloud, element, and some others).

Anyway can definitely recommend, it feels like a real cluster but it's all on a single piece of hardware which makes snapshots etc super simple.

This seems kinda backwards, as far as I know charities and donation schemes are overwhelmed by clothing donations. Our problem isn't that we are destroying clothes that could be used somewhere else, the problems are manufacturing low quality clothing that lasts 2-3 wears, and fast fashion where people buy clothing for a specific event.

If they want to achieve their goals they should be aiming for demand destruction on _new_ clothes, once the clothes are unwanted it's too late.

But seems better to somehow incentivize fabric recycling and higher quality clothes. Even expensive clothes fall apart these days.

Loved my oneplus2, the rest were mediocre at best.

Went from great value hardware with open, minimalist software to overpriced hardware and shitty bloated software.

Great example of how chasing short term wins can bleed you dry over a few years

We already have tags for PDF and Video so I could definitely see one for [LLM] working!

Generally I'd appreciate a top level comment from the submitter saying this is LLM generated but I read it and found it interesting because x,y,z because I'd rather read slop someone vouches for than slop someone hasn't

because if you don’t put this disclaimer the top comment is always "Acthually this isn't real science because you didn't publish your P value" so you can't win.

also the article itself is clearly LLM generated though

This is a scene in Catch22 IIRC where they decide to stop training the soldiers who need to parachute behind enemy lines because the fatality rate of training was so high that mathematically it made more sense to send in untrained soldiers.

non-english, twitter, and probably LLM-generated, and still on the front page. So either people really care about this issue or it's artificial as usually any one of those is enough to get ignored or flagged

Anyway here's an English translation for ease (google translate)

Wild West @dziki_zachod The EU is simultaneously working on five regulations that will change how we live, drive, and use the internet.

Here's what's currently in the legislative process—and what's already in effect.

Advanced Driving Cameras (ADDW)—in force since July 7, 2026. Every new car must have a camera tracking the driver's face and eyes. Officially for safety. The data is supposed to stay in the car. For now.

Active Speed Limiting (ISA)—in force since July 2024. New cars automatically limit their speed to the limit. It can be disabled, but the system resets itself.

Chat Control—still under negotiation. The project involves mass scanning of private messages for illegal content. Criticized by cybersecurity experts as the end of end-to-end encryption. Blocked several times, it's back in new versions.

Online KYC/age verification – just announced. Von der Leyen presented an app requiring ID to access certain platforms. The Vice-President of the European Commission announced that VPNs will also be regulated to ensure this cannot be bypassed.

Hate speech as an EU crime – The Commission wants to expand the list of EU crimes to include hate speech. Critics warn that the definition is so broad that it could criminalize simple political criticism.

Each of these regulations officially has a noble purpose. Together, they paint a picture of a Europe where you can say anything you say in front of a camera – and someone records it. 10:00 AM · Jul 8, 2026 32.6KViews

something similar actually really annoys me on linkedin mobile, I'm left handed and often accidentally like posts if I scroll my feed as the like button is very close to where I naturally touch the screen to scroll.

Yeah all claude models are doing this now. I also had a flow where it would enter username and password for demo server that are literally displayed on the page for any human to login. A couple of weeks ago claude would happily use chrome to take screenshots after logging in, now it flat out refuses and says I need to give it page where I've logged in and that it can't make an exception even if credentials are demo/demo and available to anyone to use. Super annoying stuff.

And here's a similar list of 3400 VCs who suddenly find themselves a lot less open to cold email

https://apparent.social/heat-map

Nvidia is already selling exactly this I think, not sure when it's expected to ship

I have huge respect for Armin but all of the concerns about agents producing more code with less competent supervision from senior engineers doesn't seem that different from the status quo to me. A vast majority of all software I've ever professionally worked with has been terribly structured, hard to work with, full of bugs, etc, produced by mediocre to bad engineers and run by semi technical product owners or managers who basically promot the software into existence by making jira tickets on 2 week cycles to hold it together.

Yes it's awful, but it kind of works and has worked for a very long time. Agents are already improving a lot of open source software. Yes they're producing a lot of slop too, but having beautiful code, understanding how the system works and being able to delegate to a competent engineer you trust is reserved for the very few right now and I think we have all the systems and experience in place to deal with "bad" but working software so personally I am not concerned

Some interesting stuff with a lot of LLM tells (part that actually matters, quietly, etc) which makes me skim instead of read.

Llm content can be good, but if I can't tell if it's slop or not at first glance (eg rent $800/year looked sus) I'm going to assume slop until someone I trust reads it and tells me it isnt.

I've read so many versions of this idea over the last few years and I don't understand what the point is. It's either obviously true (humans have this thing we call consciousness and by definition it's specific to humans or organic minds so therefore silicon can't have it) and therefore not worth saying or it needs to be said in some falsifiable way "an AI would be called conscious if it could fulfil conditions A, B, C".

But usually it's some long manifesto about loving a baby or experiencing hunger or something. I'm not saying I disagree or that I believe that AI is "conscious" I'm just not sure what the point of the debate is and what this word 'consciousness' (that philosophers have anyway been debating the meaning of for centuries) has to add to the AI discourse.

Enjoyed playing it, here's the direct link to play as otherwise you have to click from the article to the GitHub and then find the correct demo link

https://vnglst.github.io/when-ai-fails/shepards-dog/claude-f...

Might be flagged as slop but I asked Opus to do this and was very unimpressed by the result so when I got Fable I asked it to redo.

I need to spend some more time reading Fable's one but so far I'm quite impressed. This paragraph was also interesting

And the question this site exists to raise: why not an AI king? It is 2026; the thought is no longer science fiction, and a document written by an AI owes it a straight answer. The answer is no, and the reason is not that the AI would be stupid. It is that the proposal does not solve the power problem — it relocates it, to whoever trains, instructs, audits, and can switch off the AI, which is to say it creates the most concentrated and least transparent throne in history and then asks who holds it. Premise 5 does not care that the throne is made of silicon; it asks, as always, who corrects the error? — and premise 9 guarantees there will be errors, baked invisibly into objectives and training data, executed at machine speed. An unaccountable perfect ruler is a contradiction: unaccountable means the errors do not surface, and errors that do not surface compound (premise 9) until they are catastrophic. What AI is actually for, in utopia, is the opposite role — instrument, not throne: the auditor that reads every public contract for theft, the translator in every courtroom and clinic, the forecaster that prices the migration formula, the tireless eye that makes transparency-by-default searchable by anyone. Mechanical, inspectable, swappable, and subordinate. Take it from the AI: do not crown one.

uh they get denied entry if they are late because german trains often are and it wreaks havoc on swiss timetabling where trains still generally depart to the minute and many commuters plan their day around making connections with a 2 minute change time. if the ICE from basel to zurich is late then switzerland runs their own replacement in its spot and denies entry to the german train to avoid knockon delays.

yes the schedule is full but its not just no space for more trains, more no space for unpredictable trains

There is definitely a lot of this happening, e.g. this is a 'collaboration suite for civil servants' that's basically a collection of existing open source projects

https://github.com/MinBZK/mijn-bureau-infra/

They show all the components they use here https://minbzk.github.io/mijn-bureau-infra/docs/category/com... and have set up guides for departments to operate it all on Kubernetes

I'm guessing from my own use of NextCloud, Matrix etc that this will simply be deemed not good enough compared to Google Workspace or Microsoft WhateverItsCalledNow as these things are pretty rough around the edges in my experience, but this looks like a good step in the right direction to me

Fair enough, it's a toy project from when first nano banana came out and I wanted to see what I could do with it.

I left it running because it's my first project ever to get traction in terms of regular users, even though it costs me a few dollars a month to run. It's clearly not good enough to pay for

Some style guides distinguish between an acronym (Nasa, say the word as it's written) and initialism (ISS, say each letter)

If it's any consolation, I got 500+ people to use my similar site (a bit more niche, telegram bot that generates photoshoots from a selfie) but not a single one converted to the $5 upgrade so I just spent some money on Gemini credits for nothing.

https://flash.getgai.com/ if anyone else wants to join

The ICEs are still generally ok, as long as you're willing to play roulette with big delays and missed connections (last few I got ran perfectly).

Regional trains are part of Germany's new super cheap unlimited travel ticket, which excludes ICE and afaik this is why they are now super crowded but I haven't taken them recently so no first hand experience

Yes you can see all the working examples that people have tried just below the input.

I'm not doing anything fancy with the scraping so even if no paywall but some botblocking it will probably fail but worked across bbc, cnn, politco, and several other major news sites I tried

ah right, I think that might be what is "Telegram Lite" on the App Store

These are the two options I see to download https://i.ritzastatic.com/static/1e133ef5057a949b7ddd92e5668...

And the 'main' one that I usually use doesn't have export settings that I can find

https://i.ritzastatic.com/static/18db23448a373338766bf419fa0...

I do some similar charting etc with telegram data dumps that you can still get from the "telegram lite" desktop app even though they removed the export functionality from the main app.

For removing noise you might want to look into TF-IDF instead of the manual method described in the post that I didn't understand. It basically looks for words common across the whole corpus as noise or ones that appear within a specific chat much higher than the whole dataset as interesting.

You can also do some fun stuff by finding phrases used asymmetrically eg more by one person in the convo than the other, or over time.

Wordclouds per person are also fun!

Yeah with AI voice cloning I'd definitely be default suspicious now but at this stage I know close friends and relatives well enough to be able to ask one question to establish identity

Pmf is this weirdly defined thing where "if you're not sure you have it then you don't".

I think it was clearly useful for months to people who had tried it and taken the time to understand it, but now that knowledge has spread to the point where wallet holders are convinced it's not just passing fad or hype so now pmf can be "claimed".

I agree it's weird to say "those people have pmf" though, usually it's something you define for yourself

I did something similar, just a photo of handwritten phone mumbers and an easy to remember URL that's not indexed.

Anyone will hopefully lend you a phone if you're in a pinch but I realized that I don't know very numbers to actually call and it's kinda weird to start using email/Whatsapp whatever on a strangers phone compared to asking to visit one site and make one call

I found a GitHub that was spreading malware. I chatted to AI and it cloned it into a vm, did some static analysis, pulled the usernames of the bad guys and the guys they were impersonating and gave me a summary and an email address I could contact. The guy replied with a note of appreciation.

I had a shitty android app that I'm forced to use, I chatted to AI and it reverse engineered a binary that talks to some hardware using Bluetooth and built me a simpler version with only the buttons I needed.

I love chatting to AI.

The people the OP describes are assholes and AI amplifies them. I've met all those people and it annoys me too. The internet and phones made all of these problems worse too but we developed spam filters and trained people not to mass forward stupid meme emails.

We'll learn how to use AI better and develop better controls for people to get away from it too but it's a huge net positive from my perspective

Dutchnews.nl is an established media outlet and what a lot of English/non-Dutch speakers use to keep up with Dutch news (which is perhaps surprisingly to Americans, usually written in Dutch)